0

let us consider following data from finance.yahoo.com

http://finance.yahoo.com/q/hp?a=&b=&c=&d=8&e=16&f=2014&g=d&s=MTLA.HM%2C+&ql=1

for MOTOROLA SOLTN (MTLA.HM), i want to analyse for example its spectra; structure ,identify if it contains some periodic components, identify significant frequencies and so on, but for correct analysis it is important to know its sampling frequencies , can i identify it with this parameters

Daily
Weekly
Monthly

? thanks in advance

dato datuashvili
  • 9,194
  • I'm not sure what exactly you are asking, but if you use a sampling frequency of $f_s$, then the sampled spectrum has a range of $[-f_s/2,+f_s/2)$. Thus, if you look at the spectrum using the daily data, it will have all frequency components corresponding to daily fluctuations and slower (i.e., weekly, monthly, yearly, etc.). – AnonSubmitter85 Sep 16 '14 at 19:28

1 Answers1

1

Can't comment (insufficient reputation) - so please accept this as a form of answer.

You don't say precisely what the purpose of the analysis is. I am guessing, but it seems like you may be wanting to analyse stock movements to find patterns? I think you want to be more specific in the question - at the moment it is very broad.

If I think about the question from a trading perspective. People trade shares on the basis of a variety of sampling frequencies. If your strategy is medium term (i.e. weeks to months) holding of shares then you are likely to be looking at daily or weekly movements. Others may trade between days or even within the day (i.e. having a zero position overnight) - for those even daily data is too infrequent and they may be looking at data ever minute/few minutes. These type of traders are generally trading off the data and market sentiment rather than company fundamentals.

I would presume there is a segment of traders who are looking at monthly data - those that are trading fundamentals of the company the share represents.

The other issue you have with lower frequency data is that the longer the time period a certain number of points represents - and therefore the data probably becomes more influenced by the general financial environment.

My view would be more frequent data (daily or better) and then take a peek at weekly and monthly.

AndyL
  • 58
  • thanks for reply, if i would take daily then what sampling frequency will be good choice – dato datuashvili Sep 16 '14 at 16:08
  • purpose is spectral analysis – dato datuashvili Sep 16 '14 at 16:09
  • to show main frequencies component – dato datuashvili Sep 16 '14 at 16:11
  • If you have daily data - i.e. like the Motorola dataset then the sampling has already been fixed in the dataset. You will be limited by the dataset as to the highest frequencies you find in the data. – AndyL Sep 16 '14 at 16:29
  • if it is fixed then how can i determine it? – dato datuashvili Sep 16 '14 at 16:31
  • does sampling frequency is equal to 1? – dato datuashvili Sep 16 '14 at 16:33
  • i would like to find data that can be analysed by fourier method – dato datuashvili Sep 16 '14 at 16:36
  • Any sampled data can be analysed, however the sampling frequency will determine the highest frequencies that can be found in the data. If your data is sampled once a day you will not be able to find any frequencies higher than "per 2 days" (you need more than 2 samples per cycle). I suggest that you do a search on "frequency aliasing". – AndyL Sep 16 '14 at 17:44
  • in other word ,if i will take [pxx,f]=periodogram(mydata,[],[],samplingfrequency] what should i write for samplingfrequency? – dato datuashvili Sep 16 '14 at 18:11
  • @dato - this feels like we are moving increasingly away from a mathematics question. I don't know how you are calling periodgram and I have not used it myself, but the sampling frequency of your data is per day, per week or per month. – AndyL Sep 17 '14 at 03:32